Newshot

Unitree's 王兴兴 says robot model generalization is today's toughest problem: he wants robots to work immediately in every household without retraining

AI summary

Speaking at the 2026 APEC CEO Summit in Shenzhen on November 17, Unitree Technology Chairman 王兴兴 said robotics has grown hot in China and worldwide because people see the growing value of language and text AI models and want robots to use AI to do real work, but there are many challenges. The biggest is model generalization: he wants one model and one robot to work immediately in any household without retraining for each new scenario, and he said this year's global technological breakthroughs suggest a tipping point may be near. He also cited the strain a real AI breakthrough would put on robot manufacturing, reliability and after-sales service—scaling from tens of thousands of robots a year to millions or tens of millions—and compute demand that may far exceed most people's imagination, and noted earlier remarks that physical AI is showing signs of breaking through with GPT-6's release, with the core challenge being aligning AI model inputs and outputs with real robots and commercial services likely landing faster than home and industrial applications.

Translation(model translation, not the source's own text)

王兴兴 of Unitree says robot model generalization is the thorniest problem right now: the hope is that robots can get to work immediately in any household without retraining

IT之家, November 17 — The 2026 APEC CEO Summit is being held in Shenzhen from November 17 to 18 under the theme "Jointly Promoting Prosperity, Winning the Future Together."

王兴兴, chairman of Unitree Robotics Co., Ltd., said at the summit that in recent years the robotics industry in China and around the world has been very hot, because people have seen the growing value of pure-language or text-based AI models in personal life and office work, and they hope robots will also use AI to actually do work and solve problems on the job. But he acknowledged that there are no shortage of challenges. Among the biggest challenges today, the first is the model's generalization capability. We hope that when a robot arrives in any household, one model and one robot can do the work and do it right away, rather than having to retrain it every time the scenario changes.

王兴兴 believes generalization is the thorniest problem at the moment. Still, he said that given this year's technological breakthroughs worldwide, the tipping point for a breakthrough is already in sight, and it may come fairly soon.

王兴兴 mentioned that the question he often thinks about lately is: if robot AI models make a real breakthrough next year, what is the most important thing now and in the coming years? In his view, the challenges in this regard are also considerable — once AI models truly break through, they will put great strain on robot manufacturing, reliability and after-sales service. Right now we might produce tens of thousands of robots a year, but in the future, if we are suddenly asked to produce millions or tens of millions, that challenge is still very big.

In addition, he raised the issue of compute. I think future demand for compute may far exceed what most people imagine today. Demand for compute in the foundational layer is also very large. Including subsequent issues such as robot safety, the challenges remain very big.

As IT之家 previously reported, at the 2026 World Physical AI Conference and Unitree Partner Ecosystem Conference, 王兴兴 said that almost everyone in the world, including himself, does not attach enough importance to and belief in AI.

王兴兴 said that with the release of GPT-6, physical AI is already showing signs of being broken through, and quite possibly in the next 1-2 years or 2-3 years; the core challenge for embodied intelligence to pass its ChatGPT moment is aligning the input and output of AI models with real robots. He believes that deployment in commercial services will come faster than in the home and industrial sectors.

王兴兴 also revealed that Unitree's future focus will be on conquering robot brain technology; in addition to its own technological breakthroughs, the company will seek cooperation globally, is bullish on physical AI technology, and is committed to making breakthroughs.

Related articles

  • Why are robot combat performances less interesting than last year's? Unitree CEO 王兴兴 responds
  • 王兴兴 of Unitree: next year will most likely be the tipping point for the robotics industry or robot models
  • 王兴兴: Unitree will focus on conquering robot brain technology in the future and plans to seek global cooperation
  • 王兴兴 lays out Unitree's three major future directions, with tens of millions of yuan invested each year to host the World Physical AI Competition
  • 王兴兴 of Unitree: I'm afraid of wasting every day; this generation has the hope of lifting humanoid robot productivity to the next level
  • 王兴兴 of Unitree: physical AI is already showing signs of being broken through

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗